On Tue, 13 Feb 2001, Stephan Kulow wrote:
> > In order to improve the algorithm, Dr. Tom Holroyd (who is
> > subscribed to fc-solve-discuss) suggested using threads that will
> > operate on the same state collection. Each thread will have a different
> > tests order, or a different weight function, which will increase the
> > chance that one of them will work.
> Yes, fuck the CPU - we want threads :)
>
So do I. But I also need to program them. ;-)
> What is your current weight function? I still think that the multi-move
> approach doesn't go well with A* speed wise. You're right that this
> helps
> reducing memory usage, but it also widens your search tree
>
Hmmm. Well you can try taking Tom's solver and implement A* in it. It
already implements DFS and BFS. I can also program atomic moves tests.
I can set the relative weights of each consideration in the weight
function. So far, I've seen that for each initial board, one configuration
of weights seems to work well, but there wasn't one that worked well all
the time. Currently the parameters that are considered are: (out of USAGE)
1. The number of cards out.
2. The maximal sequence move.
3. The number of cards under sequences.
4. The length of the sequences which are found over renegade cards.
5. The depth of the board in the solution.
You can try playing with their relative weights and see if you can find
something that will solve the first 1000 Microsoft games in a shorter time
than Soft-DFS.
Regards,
Shlomi Fish
> Greetings, Stephan
>
> --
> Sitzen zwei Emanzen am Frühstückstisch. Sagt die eine zur anderen:
> "Du, gib mir doch mal bitte die Salzstreuerin!"
>
----------------------------------------------------------------------
Shlomi Fish shlomif_at_vipe.technion.ac.il
Home Page:
http://t2.technion.ac.il/~shlomif/
Home E-mail: shlomif_at_techie.com
The prefix "God Said" has the extraordinary logical property of
converting any statement that follows it into a true one.
Received on Tue Feb 13 2001 - 06:44:56 IST